I wrote a piece about Dai in which I suggested it was vulnerable to certain black swan events that I felt were not being given appropriate consideration. I do still feel that those are accurate opinions, however, I have also come to appreciate more of the ethos surrounding Dai.
Dai serves a valuable purpose. It was intended to be a censorship resistant stablecoin, which is in sharp contrast to the wide variety of stablecoins like Tether that can be easily censored.
For myself, censorship resistance is the defining feature of what makes cryptocurrencies valuable and interesting. Traditional finance and computers have already enabled or potentially enable a huge amount of automation and disintermediation, all without ever needing a blockchain. However, those still depend on a centralized entity for their survival, and for you to use them. Cryptocurrencies are meant to stand in contrast to that and provide resilient platforms that cannot be censored.
Dai is an over-collateralized stablecoin that relies on a variety of different collateral types which are deposited into ‘vaults’ to create Dai. The different types of collateral have different risk profiles which affects how much Dai you need to return to the vault to unlock your collateral. The types of collateral and these risk parameters are decided on by holders of the governance token MKR.
How does Dai stay at a dollar?
Dai stays at or above a dollar by ensuring that every single Dai is over-collateralized. Each vault that initially creates the Dai has more collateral than the number of Dai generated. If the vault gets close to not being sufficiently collateralized then the collateral is auctioned off for Dai. If instead the price of Dai trends to over a dollar then creation of more Dai needs to be incentivized. There are several methods that are used to incentivize this with the simplest being reducing the cost to create a vault (the ‘interest’). This makes it more appealing for people to lock up their collateral and receive Dai in exchange.
There have been other mechanisms implemented to help maintain the peg, each with their own added degree of complexity. One of these is the ‘Dai Savings Rate’ (DSR). The DSR is a mechanism by which someone can lock up their Dai and earn interest on it. It is also used to help maintain the peg. If you need to remove Dai from the market to restore the peg you can make it more attractive for someone to save it. If you need people to get more Dai into the marketplace then you can drop the interest rate until individuals are incentivized to withdraw their Dai and find other productive uses for it.
Another mechanism is the ‘Peg Stability Module’ (PSR). The PSR is a special type of vault. It allows for Dai to be traded directly for USDC (or vice-versa) for a small fee. Thus, if Dai is above a dollar then traders are incentivized to use this module to trade their USDC for Dai. If Dai is under a dollar then this provides an easy way to exchange your Dai for USDC with your only cost being the fee associated with the module and effectively exit Dai for fiat. This trade only works for as long as there is USDC remaining in this module.
What if the collateral crashes?
Astute readers have likely noted that while every Dai begins its life over-collateralized, many of the assets that been chosen to collateralize Dai can be somewhat volatile. In order to prevent this (nearly) every vault has a ‘liquidation ratio’ somewhere above the $1 value mark. So for example if you have an Ether vault that starts at 150% collateralized, but that has a liquidation ratio of 1.1 then if the value of your Ether falls to 110% of the amount of Dai then your vault starts being auctioned off. This is done to try to ensure that Dai stays over-collateralized.
These auctions work by getting a bid for all of the collateral. If this bid is enough to pay off the vault’s obligations with some leftover then the auction changes. It becomes a ‘Reverse Collateral Auction’ in which the protocol determines the minimum amount of collateral that needs to be sold in order to pay off the obligations. Any collateral remaining after this is returned to the creator of the vault. If insufficient Dai is raised in the auction then it becomes a debt which is paid by the ‘Maker Buffer’ which is a collection of Dai that has accrued from fees and auctions.
Now what if a whole bunch of vaults were liquidated nearly simultaneously? Well in cases where the debt cannot be covered by the ‘Maker Buffer’ then the governance token holders get diluted until the system is recapitalized. So the system issues and sells more and more of the MKR token until the system again is collateralized.
What if the collateral is seized?
When I say seized I mean frozen in the sense that USDC or USDT (two collateral types for Dai) can be easily frozen by their central issuer. So what happens if these tokens are frozen while they are in a vault? I am not 100% certain but will give my best explanation I have been able to determine so far. By default it seems to have no effect. Namely those vaults will still exist and will still contain those tokens, but those tokens will now no longer be able to be transferred. This means that vault is nearly guaranteed to be abandoned, as the original creator will have no desire to get their collateral out.
If this happened on a large enough scale then you might start to see significant dilution of the MKR token as auctions for these collateral are unlikely to raise enough to pay their obligations.
What problems does this lead to?
There are several potential problems with the current construction of Dai/MKR. First let’s discuss the seizure risk. At times nearly half of the Dai collateral has been in the form of these easily freezable/seizable assets.
This is even further complicated by the fact that a significant portion of the Dai collateral is in the form of USDC-A which is a special type of vault that….cannot be liquidated. If the value of USDC deviates from a dollar or these USDC are blacklisted then that entire section of the Dai collateral can now not be recovered.
Further complicating this is the ‘Peg Stability Module’. This module potentially allows for significant accumulation of stablecoin backing in times when Dai is not at its peg.
This means that there are certain black swan events wherein a significant portion of the vaults are liquidated, while many that should be liquidated are not, leaving Dai in a severely under-collateralized state.
Compounding this risk is what is likely to be happening to the broader market if one of these ‘stable’ assets is breaking peg or freezing a huge number of coins, which I would expect to be a market downturn. Since much of the non-stablecoin backing of Dai is in highly correlated crypto-assets they can further compound this under-collateralization.
Thus as those assets move down their vaults need to be liquidated. These liquidations can end up pouring gasoline on the fire. Namely that as more of this collateral is being released into a market going down it can accelerate the rate that more vaults are liquidated. This feedback mechanism gets more powerful the larger that Dai becomes.
The more that Dai depends on other stablecoins the more vulnerable it becomes to black swan events.
How would I improve it?
Fundamentally, to build a censorship resistant and resilient system you must ensure that you are integrating components that are censorship resistant themselves.
So if I was a MKR holder I would be doing everything in my power to ensure that no more easily seizable/freezable assets are added to collateral. Then I would start adjusting the risk protocol for ones that have already been accepted as collateral to make it increasingly unattractive to create vaults using these types of asset. Ideally these assets never should have been accepted as collateral.
I would entirely eliminate the peg stability module. It was a decision that traded long term resiliency and censorship-resistance for peg stability. Dai will likely always be a ‘soft’ peg and to maintain its resiliency against black swan events it is better to allow it to break the peg (to a degree) than it is to take on this additional risk by accepting this kind of collateral.
If I was a god with endless power I would also try to eliminate the influence of the MakerFoundation and the early investors in MKR. Having a governance structure at all increases your attack surface and having easily identified individuals with significant influence over the protocol makes it significantly riskier.
I am tempted by the idea that the only collateral should be Ether, but I also do think there is value to diversification. However, currently the value of the diversification is limited due to the overall correlation of the market.
Summary and Conclusion
Overall I think Dai and MKR are incredibly clever. The idea of a censorship resistance stablecoin is quite fascinating to me. However, I worry that decisions that were made to help maintain the peg and increase the use of Dai have resulted in it being much more vulnerable and much less censorship resistant.
Originally published at http://bennettftomlin.com on March 11, 2021.